Workload, fatigue and the vigilance decrement: the cost of sustained attention and the limits of the laboratory number

Vigilance and fatigue research supply this module's few quantified laws. Watching for a rare signal degrades with time on task, and wakefulness impairs performance dose-dependently.

Both are real, and their reach is short. The workload score is a weighted judgement, and self-report saturates before impairment does. The work system produces ward workload. This fourth of ten units measures what sustained attention costs, and finds the laboratory law shorter than its applications assume.

The decrement was measured before it was explained

Mackworth's study began as an operational problem: radar operators were missing returns their equipment presented. He reproduced the watch in the laboratory with a clock face whose pointer occasionally jumped two steps, over two hours (Mackworth, 1948). Detections fell, most steeply in the first half hour.

See and colleagues pooled 42 vigilance studies covering 138 experimental conditions (See et al., 1995). They were testing the taxonomy that attributes the decline to discrimination type and event rate. They confirmed a substantial sensitivity decrement, but they showed its magnitude depended on both. The effect size is therefore a function of the task.

Warm and colleagues reviewed task-type comparisons, transcranial Doppler measures of resource demand and task-induced stress (Warm et al., 2008). Behavioural, neural and subjective findings converged on vigilance as hard mental work.

The decrement's mechanism and measure are both disputed

Thomson and colleagues judged neither the monotony account nor resource depletion adequate alone (Thomson et al., 2015). They propose a resource-control account, in which attention lapses as executive control is withdrawn.

Measuring a sensitivity loss needs accurate hit and false-alarm rates, and typical vigilance tasks induce false alarms so rarely that the analyst cannot distinguish the two (Skinner et al., 2023). The authors raised false-alarm rates by overlapping the target and non-target distributions, and the decrement was still not reducible to sensitivity. A ward's false-alarm policy moves exactly that criterion.

Workload became a number, and the number is a rating

Hart and Staveland developed the NASA Task Load Index (Hart and Staveland, 1988). It rates six subscales, mental, physical and temporal demand, performance, effort and frustration, and weights them into a global score. Every subscale is a subjective rating, so scores from different settings compare raters as much as tasks.

Hart's own survey for the twentieth anniversary reviewed 550 studies that had used the task load index (Hart, 2006), and found its use had spread beyond aviation, beyond crew complement and beyond English. The subscales are still read by the rater, so a ward reading them in a second language needs the wording checked. A global score carries its meaning only against the population of scores it belongs to.

Wickens set out multiple resource theory and its four-dimensional model (Wickens, 2008). Interference rises with the sharing of processing stage, sensory modality, code and visual channel. Predictions correlated highly with observed interference in driving simulation, though lower correlations identified resources the model lacks.

Young and colleagues found mental workload among the most used and most nebulous concepts in ergonomics (Young et al., 2014). They name quantification of workload red lines as a contemporary challenge. A red line is the threshold at which operators approach their performance tolerances. A quantity with no single referent is not yet a design law.

Performance can look stable while the operator pays

Hockey's cognitive-energetical framework explains why one output measure misleads (Hockey, 1997). Goals and actions are regulated by a compensatory control mechanism that allocates resources dynamically. Performance under stress may be protected by recruiting further resources, at the expense of increased subjective effort and physiological costs. It may instead be stabilised by lowering the goals. With no primary-task decrement detectable, the framework predicts disrupted subsidiary tasks and strain.

Nothing going wrong has measured the primary task only. The effort spent and the tasks dropped are invisible to an outcome count.

Fatigue: the law with the widest reach

Dawson and Reid kept 40 subjects awake, in one condition for 28 hours (Dawson and Reid, 1997). In a counterbalanced condition they raised blood alcohol to 0.10 per cent and measured tracking half-hourly. Between the tenth and twenty-sixth hours mean relative performance fell by 0.74 per cent per hour. Each 0.01 per cent rise in blood alcohol cost 1.16 per cent. 17 hours awake was equivalent to 0.05 per cent blood alcohol, and 24 hours to about 0.10 per cent.

Van Dongen and colleagues randomised 48 healthy adults to 4, 6 or 8 hours in bed for 14 nights, or three nights without sleep (Van Dongen et al., 2003). Restriction to 6 hours or less produced dose-dependent deficits equivalent to up to two nights of total sleep deprivation. Sleepiness ratings rose acutely, barely moved thereafter, and failed to separate the 6-hour and 4-hour conditions. Lapses related near-linearly to cumulative wakefulness beyond 15.84 hours.

Akerstedt and Gillberg recorded EEG and EOG continuously against sleepiness ratings every two hours, in eight subjects kept awake in a time-isolated laboratory (Akerstedt and Gillberg, 1990). Subjective and objective sleepiness covaried closely but curvilinearly. Physiological signs appeared reliably only once self-rated sleepiness was high on the nine-point Karolinska Sleepiness Scale.

The laboratory law reaches the ward

Gurses, Carayon and Wall surveyed 265 nurses in 17 intensive care units, an 80 per cent response rate (Gurses, Carayon and Wall, 2009). They modelled performance obstacles, work-system factors that hinder a nurse's capacity to do the job, against workload and perceived quality and safety of care. Obstacles raised workload, and workload mediated their effect. Equipment-related obstacles acted without mediation.

The fatigue relation tells a designer how much a tracking task degrades per hour of wakefulness. It does not tell a manager what a night shift at reduced staffing does to a ward. The ward also staffs older nurses and people holding a second job. There the workload is produced by obstacles a designer can move, and it is partly compensated by effort that no outcome count sees.

The contested question: depletion, control or criterion?

Three positions are held on what the decrement is. The resource position holds that sustained attention consumes a limited resource whose decline reduces performance. Its evidence is converging behavioural, neural and subjective findings, with the successive-simultaneous contrast supporting the account (Warm et al., 2008). The control position holds that the lapse is a withdrawal of executive control that lets mind wandering take over (Thomson et al., 2015). The criterion position holds that the standard measure cannot separate a sensitivity loss from a criterion shift (Skinner et al., 2023). Each names a test: a resource index tracking within-person change, an independent measure of mind wandering, and false-alarm rates high enough to falsify a sensitivity account.

One side can point to laws holding across conditions within an experiment. The cumulative-wakefulness model predicted lapses across all four sleep conditions of its study (Van Dongen et al., 2003). The other can point to four gaps. A pooled effect has a size that depends on the task (See et al., 1995). An instrument returns a score that is a weighted judgement (Hart and Staveland, 1988). A construct has no agreed thresholds (Young et al., 2014). The work system produces ward workload (Gurses, Carayon and Wall, 2009). A prospective ward study changing one named obstacle or shift length would settle it. Such a study would test the prediction against a performance measure and a self-report.

What to do while the dispute is open

The evidence on the two sides is unequal. The laboratory laws rest on controlled studies with large effects that replicate. The transfer claim rests on one cross-sectional survey of nurses and on instruments that return weighted judgements. The strongest objection to this unit's thesis is that a cross-sectional ward survey cannot show that a law demonstrated under control fails to travel. While the dispute is open, a designer should measure effort and the tasks dropped, not the output alone, and change a shift length or a named obstacle rather than treat a workload score as a threshold.

Where this is heading

Two directions are visible. First, the outcome measure of vigilance is being reopened. If a low-signal task cannot separate a sensitivity loss from a criterion shift, the next claims will rest on tasks with measurable false-alarm rates, and Skinner and colleagues (2023) set that standard. Second, workload is moving from a global rating towards red lines and structural prediction. Young and colleagues (2014) name tolerance thresholds as the open problem, and Wickens (2008) offers a resource model built to predict overload before it occurs. The thread for health is Gurses, Carayon and Wall (2009), whose obstacles a trial can change.

Sources

• Mackworth, N. H. (1948). The Breakdown of Vigilance during Prolonged Visual Search. Quarterly Journal of Experimental Psychology. doi:10.1080/17470214808416738 Reproduced the radar operator's watch as the clock test, a pointer stepping in regular increments and occasionally jumping two, over a two-hour vigil. Detections declined, most steeply in the first half hour.
• See, J. E., Howe, S. R., Warm, J. S. and Dember, W. N. (1995). Meta-analysis of the sensitivity decrement in vigilance. Psychological Bulletin. doi:10.1037/0033-2909.117.2.230 Pooled 42 vigilance studies covering 138 conditions to test the taxonomy attributing the decline to discrimination type and event rate. It confirmed a substantial sensitivity decrement whose magnitude depended on both, so the effect size is a function of the task.
• Warm, J. S., Parasuraman, R. and Matthews, G. (2008). Vigilance Requires Hard Mental Work and Is Stressful. Human Factors. doi:10.1518/001872008x312152 Reviewed task-type comparisons, perceived workload, transcranial Doppler measures of resource demand and task-induced stress. Findings converged on vigilance demanding hard mental work.
• Grier, R. A., Warm, J. S., Dember, W. N., Matthews, G., Galinsky, T. L., Szalma, J. L. and Parasuraman, R. (2003). The Vigilance Decrement Reflects Limitations in Effortful Attention, Not Mindlessness. Human Factors. doi:10.1518/hfes.45.3.349.27253 Tested a traditional vigilance task against a routinised version built to promote mindlessness, and found the decrement in both. It is the primary experiment behind the effortful-attention account the unit takes from a review.
• Thomson, D. R., Besner, D. and Smilek, D. (2015). A Resource-Control Account of Sustained Attention. Perspectives on Psychological Science. doi:10.1177/1745691614556681 Reviewed the sustained-attention literature against the monotony and resource-depletion accounts and judged neither adequate alone. It proposes a resource-control account combining attentional-resource and control-failure theories of mind wandering.
• Skinner, H. E., Ruacho, I. and Giesbrecht, B. (2023). The Vigilance Decrement is Not Only About Sensitivity. Journal of Vision. doi:10.1167/jov.23.9.5828 Raised false-alarm rates by manipulating target and non-target distribution overlap, noting that a sensitivity loss can be measured only with a usable false-alarm rate. The decrement was not reducible to sensitivity and involved a criterion shift.
• Hart, S. G. and Staveland, L. E. (1988). Development of NASA-TLX (Task Load Index): Results of Empirical and Theoretical Research. Advances in Psychology. doi:10.1016/s0166-4115(08)62386-9 Reports the work behind NASA-TLX, six subscales covering mental, physical and temporal demand, performance, effort and frustration, combined into a global score by pairwise weighting. The weighting is why the number is a judgement.
• Wickens, C. D. (2008). Multiple Resources and Mental Workload. Human Factors. doi:10.1518/001872008x288394 Sets out multiple resource theory and its four-dimensional model, processing stage, sensory modality, code and visual channel, applying it to multitask driving simulation. Predictions correlated highly with observed interference, and lower correlations identified resources not yet modelled.
• Young, M. S., Brookhuis, K. A., Wickens, C. D. and Hancock, P. A. (2014). State of science: mental workload in ergonomics. Ergonomics. doi:10.1080/00140139.2014.956151 Reviews three decades of mental workload research, covering how the concept is understood, measured and applied in safety-critical systems. It finds the concept among the most used and most nebulous in the field, and names quantification of workload red lines as the contemporary challenge.
• Hockey, G. R. J. (1997). Compensatory control in the regulation of human performance under stress and high workload: A cognitive-energetical framework. Biological Psychology. doi:10.1016/s0301-0511(96)05223-4 Presents a two-level compensatory control mechanism allocating resources through an effort monitor and a supervisory controller. Performance can be protected by recruiting further resources at the cost of subjective effort and physiological costs, or stabilised by lowering goals. Stable output can hide a rising cost.
• Dawson, D. and Reid, K. (1997). Fatigue, alcohol and performance impairment. Nature. doi:10.1038/40775 Kept 40 subjects awake for 28 hours and, in a counterbalanced condition, raised blood alcohol to 0.10 per cent, measuring tracking half-hourly. Performance fell 0.74 per cent per hour between the tenth and twenty-sixth hours of wakefulness; 17 hours awake equalled 0.05 per cent blood alcohol and 24 hours about 0.10 per cent. The law rests on one tracking task.
• Van Dongen, H. P. A., Maislin, G., Mullington, J. M. and Dinges, D. F. (2003). The Cumulative Cost of Additional Wakefulness: Dose-Response Effects on Neurobehavioral Functions and Sleep Physiology From Chronic Sleep Restriction and Total Sleep Deprivation. Sleep. doi:10.1093/sleep/26.2.117 Randomised 48 healthy adults to 4, 6 or 8 hours in bed for 14 nights, or three nights without sleep. Restriction to 6 hours or less produced dose-dependent deficits equivalent to up to two nights of total sleep deprivation, and ratings did not separate the 6-hour and 4-hour conditions.
• Akerstedt, T. and Gillberg, M. (1990). Subjective and Objective Sleepiness in the Active Individual. International Journal of Neuroscience. doi:10.3109/00207459008994241 Recorded ambulatory EEG and EOG continuously against sleepiness ratings every two hours in eight subjects kept awake overnight in a time-isolated laboratory. Subjective and objective sleepiness covaried closely but curvilinearly, and physiological signs appeared only once self-rated sleepiness was high on the nine-point scale that became the Karolinska Sleepiness Scale.
• Gurses, A. P., Carayon, P. and Wall, M. (2009). Impact of Performance Obstacles on Intensive Care Nurses' Workload, Perceived Quality and Safety of Care, and Quality of Working Life. Health Services Research. doi:10.1111/j.1475-6773.2008.00934.x Surveyed 265 nurses in 17 intensive care units with an 80 per cent response rate, modelling obstacles, workload, perceived quality and safety of care and working life. Obstacles raised workload, workload mediated the effect, and equipment-related obstacles acted without that mediation. Ward workload sits in the work system.
• Hart, S. G. (2006). Nasa-Task Load Index (NASA-TLX); 20 Years Later. Proceedings of the Human Factors and Ergonomics Society Annual Meeting. doi:10.1177/154193120605000909 Surveyed 550 studies that used or reviewed the task load index, and found its use had spread beyond aviation, beyond crew complement and beyond English. It describes the settings the score has been read in, and sets no threshold for it.